The first large scale geotechnical centrifuge in Korea has recently been developed at KAIST under the Korea Construction Engineering Development (KOCED) Collaboratory program. A 5 m platform radius, 240 g-tons state-of-the-art geotechnical centrifuge has been installed in a new facility. The centrifuge has the unique feature of an automatic balancing system and includes parts for general testing purposes such as fluid rotary joints, slip rings, a fiber optic rotary joint and an Ethernet network system. In addition, a four degree-of-freedom in-flight robot can be equipped to simulate complex construction or in-situ testing process during centrifuge flight. In order to simulate earthquake motion during operation, a self-balancing type biaxial shaking table has also been developed. Since the KOCED program promotes collaboration and remote use, tele-presence and tele-participation environments have been implemented in this facility.

Simultaneous leakage of soil particles and groundwater into underground vacant spaces caused by sewer pipe defects and cracks can cause underground cavities and even ground collapse on the surface. Numerical studies of this problem are limited, and the mechanism of ground subsidence caused by leakage of soil and groundwater into construction sites has yet to be verified. In this study, the factors affecting the mechanism of ground subsidence are verified via physical modelling of a sewer pipe defect. Through visual inspection and particle image velocimetry analysis, it has been found that the groundwater flow direction, hydraulic gradient around the leakage point, and strength of the ground to support itself are the main factors that dominate the mechanism of ground subsidence. These factors can affect ground deformation and cavity expansion direction, ground subsidence development and collapse speed, and occurrence of sudden ground collapse, respectively. It was also found that pore pressure fluctuates when cavity cavern collapse and cavity expansion occur.

This paper describes some details of a self-balanced earthquake simulator on the centrifuge in KAIST and results of a series of proof tests for verifying its dynamic performance and excitation capacity. The main feature of the earthquake simulator is the dynamic self-balancing technique adopted to eliminate a large portion of the undesired reaction forces and vibrations transmitted to the centrifuge main body. This feature is achieved by embarking counter-weight platform and two back-to-back hydraulic bearings. The maximum base shaking acceleration of the earthquake simulator is 20 g in horizontal direction under 40 g of centrifuge acceleration with a maximum payload of 700 kg, corresponding to 0.5 g of horizontal shaking acceleration in the prototype scale. The loading frequency ranges from 40 Hz to 200 Hz (300 Hz) for sinusoidal (real earthquake) inputs. The dimension of slip table is 670 mm × 670 m in length and width. The proof test results show that the earthquake simulator can reproduce mono-frequency sinusoidal inputs in a wide band of frequencies as well as multi-frequency real earthquake inputs at the bottom of soil models with satisfactory fidelity, and the dynamic self-balancing contributes to the safety of the centrifuge structure.

Miniature cone penetration tests were conducted in centrifuge models to investigate the effects of various testing conditions on the tip resistance including the particle size, centrifugal acceleration related to stress level and prototype cone diameter, container wall boundary, and penetration rate. Two sand materials were selected: (1) Saemangeum and (2) silica sands. The former is natural sand with high fine contents and the latter is clean sand. A series of penetration tests was performed in six saturated soil models using an in-flight robot. Three Saemangeum sand models were prepared by means of the moist compaction method. The silica sand models were made by the air-pluviation method. Modeling of models was adopted to investigate the particle size effect using 7-, 10-, and 13-mm-diameter cones. The centrifugal acceleration effect also was studied by comparing the tip resistance profiles obtained at different g-levels using the 10-mm-diameter miniature cone. The results indicated that the particle size effect was negligible for both sands using 7- to 13-mm-diameter cones. However, the tip resistance decreased with increasing g-level at a shallower depth than a given critical depth ($D_{cr}$ ), especially for dense sand. Then, it merged to a single value at a deeper depth than $D_{cr}$ . The $D_{cr}$ was affected by g-level and soil density. Finally, an empirical correlation to estimate the soil density from the tip resistance in the centrifuge was proposed.